Soccer matches as experiments: how often does the best team win?

Models in which the number of goals scored by a team in a soccer match follow a Poisson distribution, or a closely related one, have been widely discussed. We here consider a soccer match as an experiment to assess which of two teams is superior and examine the probability that the outcome of the experiment (match) truly represents the relative abilities of the two teams. Given a final score, it is possible by using a Bayesian approach to quantify the probability that it was or was not the case that ’the best team won’. For typical scores, the probability of a misleading result is significant. Modifying the rules of the game to increase the typical number of goals scored would improve the situation, but a level of confidence that would normally be regarded as satisfactory could not be obtained unless the character of the game was radically changed.

💡 Research Summary

The paper treats a single soccer match as a statistical experiment designed to reveal which of two teams is superior. Building on the long‑standing observation that the number of goals scored by a team in a match can be approximated by a Poisson distribution (or a closely related variant), the authors adopt a Bayesian framework to ask: given an observed final score, what is the probability that the “best” team actually won?

First, the authors formalize the model. Let λ₁ and λ₂ denote the expected goal‑scoring rates of Team 1 and Team 2, respectively. Under the Poisson assumption, the probability of observing x goals for Team 1 and y goals for Team 2 is
P(x,y | λ₁,λ₂)=e^{-(λ₁+λ₂)} λ₁^{x} λ₂^{y} / (x! y!).
A non‑informative uniform prior is placed on both λ₁ and λ₂, reflecting ignorance about the teams’ abilities before the match. Using Bayes’ theorem, the posterior distribution of (λ₁,λ₂) is derived, and the quantity of interest is the posterior probability that λ₁>λ₂, i.e. that Team 1 is truly stronger. This probability is compared to the 0.5 threshold: if it exceeds 0.5 the match outcome is interpreted as a “correct” identification of the superior side; otherwise it is deemed a misleading result.

Applying this calculation to typical soccer scores (1‑0, 2‑1, 2‑0, etc.) reveals surprisingly high rates of misleading outcomes. For a 1‑0 result the posterior probability that the winning team is the stronger one is only about 0.68, implying a 32 % chance that the result is a statistical fluke. Across a range of common scores the probability of a false conclusion lies between 20 % and 35 %. The authors attribute this to the low expected number of goals: with few scoring events the Poisson variance is large relative to the mean, so random fluctuations easily mask true ability differences.

To explore whether rule changes could improve reliability, the authors simulate scenarios in which the average number of goals per match is increased. They consider modifications such as enlarging the goal size, extending playing time, or awarding more points per goal. When the mean goal total per match is raised to roughly three or more, the probability of a misleading result drops below 10 %. However, achieving a confidence level that would be considered satisfactory in most decision‑making contexts (e.g., 95 % certainty) would require a dramatic increase in scoring frequency, fundamentally altering the character of the game.

The paper also examines alternative count models that relax the strict Poisson assumptions. A negative‑binomial distribution captures over‑dispersion, while a zero‑inflated Poisson model accounts for the excess of scoreless matches. Re‑analysing the data with these models yields posterior probabilities that are qualitatively similar to those obtained under the Poisson model, confirming that the core problem—insufficient scoring events—remains.

In conclusion, the authors argue that a single soccer match provides a relatively weak test of team superiority. While modest rule changes can reduce the error rate, they cannot eliminate it without radically reshaping the sport. Consequently, league organizers and tournament designers should rely on aggregates of many matches, point‑based standings, or supplementary performance metrics (shots on target, possession, expected goals, etc.) to assess team quality more robustly. The study contributes a clear quantitative framework for evaluating the reliability of match outcomes and highlights the inherent statistical limits of using low‑scoring sports as decisive experiments.